Google will remove autocomplete suggestions targeted at elections, voters or bias towards any political candidate. Autocomplete feature helps complete a search quickly.
The search giant has also taken measures to improve the quality and reliability of its search results and news products.
“We expanded our autocomplete policies related to elections, and we will remove predictions that could be interpreted as claims for or against any candidate or political party,” said Pandu Nayak, Vice President, Search at Google.
The US-based company will also remove predictions that could be interpreted as a claim about participation in the election, like statements about voting methods, requirements, status of voting locations, and legitimacy of electoral processes.
This means predictions like “you can vote by phone”, “you can't vote by phone”, or “donate to” any party or candidate, will not appear in autocomplete.
Google confirmed that they have long-standing policies to protect against hateful and inappropriate predictions from appearing in autocomplete.
“We design our systems to approximate those policies automatically, and have improved our automated systems to not show predictions if we detect that the query may not lead to reliable content,” Nayak said.
The technology company has created an Intelligence desk which can actively monitor and identify potential information threats.
Intelligence Desk is a global team of analysts monitoring news events 24/7 that include natural disasters and crisis, breaking news moments and the recent developments in ongoing topics like COVID.
As the events occur, Google’s analysts collect data about how their systems are responding and compile reports about narratives like new claims about COVID treatments.
The product teams then use these data sets and reports from the Intelligence Desk to conduct more robust quality tests, ensuring the efficiency of the systems while people search for a wide range of topics.
Google claimed that it is now more accurate and quick to identify breaking news moments. They estimated that the detection time improved to just a few minutes from the news breaking, up from 40 minutes only a few years ago.
The Alphabet-owned company said that it will continue to train and test their systems to make the searching more reliable, avoid low-quality information and detect areas that may be at risk for misinformation.
Its improved systems can now detect 99 percent of vandalism instances on Wikipedia. It uses machine learning tools paired with human oversight to identify and address vandalism. Most vandalism on Wikipedia is reverted within minutes.
To complement Wikipedia’s systems, it has also added additional protection and detection systems to prevent inaccurate information in knowledge panels, which give a quick access to facts across the web. The data comes from hundreds of sources of which Wikipedia is one of the most comprehensive knowledge bases.
